Stem Cell Factor Is Selectively Secreted by Arterial Endothelial Cells in Bone Marrow

ARTICLE

DOI: 10.1038/s41467-018-04726-3 OPEN Stem cell factor is selectively secreted by arterial endothelial cells in bone marrow

Chunliang Xu1,2, Xin Gao1,2, Qiaozhi Wei1,2, Fumio Nakahara 1,2, Samuel E. Zimmerman3,4, Jessica Mar3,4 & Paul S. Frenette 1,2,5

Endothelial cells (ECs) contribute to haematopoietic stem cell (HSC) maintenance in bone marrow, but the differential contributions of EC subtypes remain unknown, owing to the lack

1234567890():,; of methods to separate with high purity arterial endothelial cells (AECs) from sinusoidal endothelial cells (SECs). Here we show that the combination of podoplanin (PDPN) and Sca-1 expression distinguishes AECs (CD45− Ter119− Sca-1bright PDPN−) from SECs (CD45− Ter119− Sca-1dim PDPN+). PDPN can be substituted for antibodies against the adhesion molecules ICAM1 or E-selectin. Unexpectedly, prospective isolation reveals that AECs secrete nearly all detectable EC-derived stem cell factors (SCF). Genetic deletion of Scf in AECs, but not SECs, signiﬁcantly reduced functional HSCs. Lineage-tracing analyses suggest that AECs and SECs self-regenerate independently after severe genotoxic insults, indicating the per- sistence of, and recovery from, radio-resistant pre-speciﬁed EC precursors. AEC-derived SCF also promotes HSC recovery after myeloablation. These results thus uncover heterogeneity in the contribution of ECs in stem cell niches.

1 The Ruth L. and David S. Gottesman Institute for Stem Cell and Regenerative Medicine Research, Albert Einstein College of Medicine, New York, NY 10461, USA. 2 Department of Cell Biology, Albert Einstein College of Medicine, New York, NY 10461, USA. 3 Department of Systems and Computational Biology, Albert Einstein College of Medicine, New York, NY 10461, USA. 4 Department of Epidemiology and Population Health, Albert Einstein College of Medicine, New York, NY 10461, USA. 5 Department of Medicine, Albert Einstein College of Medicine, New York, NY 10461, USA. Correspondence and requests for materials should be addressed to P.S.F. (email: [email protected])

NATURE COMMUNICATIONS | (2018) 9:2449 | DOI: 10.1038/s41467-018-04726-3 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04726-3

aematopoietic stem cells (HSCs), at the top of the hae- Sca-1bright but cannot be cleanly separated by FACS because SECs Hmatopoietic cell hierarchy, give rise to all mature hae- also express Sca-1 (Fig. 1b). The difference in staining patterns for matopoietic cells throughout life. HSCs are thought to be Sca-1 between the classical in vitro or the physiological in vivo maintained in specific niches, allowing their maintenance and staining methods implies that a sizable fraction (~20%; compare – + regulation of their fate1 3. Staining of endogenous HSCs using Supplementary Fig. 1a and Fig. 1a) of in vitro-stained CD31 CD150, CD48, CD41, and lineage expression has revealed that cells are not bona fide ECs. they are broadly distributed close to sinusoidal endothelial cells Previous studies have suggested that the lymphatic vessel (SECs)4. Subsequent studies have revealed that perivascular marker VEGFR3 can label sinusoids by immunofluorescence stromal cells enriched in mesenchymal stem cell (MSC) activity, imaging17, but commercial anti-VEGFR3 antibodies cannot marked by Nestin-GFP, expressed high levels of the major HSC distinguish SECs from AECs by FACS. Searching to fractionate + niche factors and were associated with HSCs in bone marrow the Sca-1 ECs, we used another lymphatic EC marker, PDPN18, − − + (BM)5. Selective diphtheria toxin-induced depletion of these cells which separated CD45 Ter119 CD31 cells into three distinct − + − prevented homing of progenitors in BM5, whereas the depletion populations: Sca-1bright PDPN , PDPN Sca-1dim, and PDPN − of overlapping CXCL12-abundant reticular (CAR) cells reduced Sca-1 double-negative cells (Fig. 1c). PDPN is mucin-type HSC numbers in BM6. BM HSCs were also reduced after con- transmembrane glycoprotein expressed on lymphatic ECs and ditional deletion of Cxcl12 or stem cell factor (Scf) in perivascular fibroblastic reticular cells (FRCs)19. To evaluate the spatial – + + cells7 10. Nestin-GFP stromal cells are heterogeneous in that the localisation of PDPN vessels in the BM, we injected Nestin- + NG2 peri-arteriolar fraction was found to regulate HSC pro- GFP mice (which brightly marks arteries) with fluorescently- liferation11, and contribute to BM CXCL12 content10. Con- labelled antibodies against PDPN, CD31, and VE-cadherin and stitutive deletion of Scf in endothelial cells (ECs) also reduced imaged whole-mount sternal BM. We found that anti-PDPN, but HSC numbers in BM, suggesting that ECs contribute to the HSC not an isotype-matched antibody, exclusively labelled SECs, and niche. Co-deletion of Scf in perivascular cells (Lepr-Cre) and ECs spared Nestin-GFPbright arteries (Fig. 1d and Supplementary − (Tie2-Cre) further reduced HSC numbers, suggesting additive Fig. 1b). These results suggest that Sca-1bright PDPN cells are + − and multicellular contributions in HSC niches7. AECs and PDPN Sca-1dim cells are SECs, whereas PDPN Sca- − + Whether SCF is homogenously secreted by ECs in the BM 1 double-negative cells (in the CD31 fraction) may not be ECs. remains unclear because there is thus far no method to separate To evaluate further this issue, we stained ECs in vivo by + precisely EC subtypes12. Endomucin CD31Hi (Type H) ECs injection of fluorescently-labelled Isolectin B4 (lectin from have been identified, near the growth plate and endosteum of Griffonia simplicifolia) and anti-VE-cadherin. Both methods young mice, in vessels bridging sinusoids and arterioles where similarly stained the vascular network following intravenous – they regulate osteogenesis in a Notch-dependent manner13 15. injection (i.v.) without staining of the BM parenchyma (Supple- Sca-1 antigen has been used to identify arterioles in imaging mentary Fig. 1d). We then evaluated by FACS analysis PDPN and + + analyses of the BM11,16, but whether Sca-1 can be used by itself to Sca-1 expression on Isolectin B4 VE-cadherin ECs extracted − isolate AECs by fluorescence-activated cell sorting (FACS) is from flushed enzymatically dissociated BM. Sca-1bright PDPN unknown. cells represented 8–9% of ECs identified by Isolectin B4 or VE- + Here, we report that the combination of podoplanin (PDPN) cadherin, whereas ~86% of ECs were in the PDPN Sca-1dim gate and Sca-1 expression can prospectively distinguish BM AECs (Fig. 1e). In vivo staining with either Isolectin or VE-cadherin did − − from BM SECs. We show that EC-derived SCF is exclusively not reveal PDPN Sca-1 double-negative cells, suggesting that produced by the arterial vasculature of the BM and that selective all BM ECs indeed express either PDPN or/and Sca-1. To confirm Scf deletion in AECs, but not SECs, alters BM HSC numbers. further this issue, we stained enzymatically dissociated BM with AEC-derived SCF also promotes HSC recovery after myeloabla- antibodies against CD45, Ter119, CD31, PDPN, and Sca-1 from tion. Furthermore, we demonstrate using lineage tracing that the mice that had been injected intravenously with fluorescently- − regeneration of the BM vasculature after myeloablation, is labelled Isolectin and anti-VE-cadherin. Upon gating on CD45 − + − − accomplished independently by pre-specified arterial and sinu- Ter119 CD31 cells, we found that PDPN Sca-1 double- soidal radio-resistant precursors. negative cells were not labelled by i.v. injected endothelial − markers, whereas virtually all AECs (Sca-1bright PDPN ) and + SECs (PDPN Sca-1dim) were labelled (Fig. 1f). The majority of + − Results SECs (PDPN Sca-1dim), but not AECs (Sca-1bright PDPN ), Separation of arterial and sinusoid ECs with PDPN and Sca-1. expressed EPHB4, a receptor tyrosine kinase recently described BM ECs are commonly identified as CD31-expressing cells on BM SECs by immunofluorescence staining (Supplementary − − + among the non-haematopoietic CD45 Ter119 fraction. Sca-1 Fig. 1c)20. These results confirm that the ex vivo-stained CD31 expression was previously shown to mark the arterial vasculature fraction contains non-ECs and suggest that PDPN and Sca-1 can by confocal immunofluorescence analyses of the BM11,16.To be used to study prospectively purified populations of AECs and evaluate the ability of Sca-1 expression to isolate prospectively SECs. arterial endothelial cells (AECs), we stained flushed BM nucleated cells with antibodies against CD45, Ter119, CD31, and Sca-1. − FACS analyses revealed that the vast majority (~80%) of CD45 AECs express high levels of canonical niche factors.InNestin- − + Ter119 CD31 cells co-expressed Sca-1 (Supplementary GFP transgenic mice, AECs can be further separated as Nestin- Fig. 1a), suggesting that Sca-1 expression was not restricted to GFPbright and Nestin-GFPdim fractions, whereas SECs were uni- AECs, which should comprise a minor fraction of total BM formly Nestin-GFPdim (Fig. 2a). We next isolated these three ECs11. In vivo injection of antibodies to physiologically labelled populations of ECs (Nestin-GFPbright AECs, Nestin-GFPdim ECs (anti-CD31, anti-VE-cadherin, and anti-Sca-1) revealed, by AECs, and SECs) to interrogate their transcriptome by RNA-seq + + contrast, that virtually all CD31 VE-cadherin (CD144) ECs analysis. Principal component analysis revealed that Nestin- (~99.4%) expressed Sca-1 (Fig. 1a). Whole-mount immuno- GFPbright AECs and Nestin-GFPdim AECs clustered together, fluorescence analysis of sternal bones of the same mice revealed away from SECs (Fig. 2b). Hierarchical clustering analysis of uniform labelling of the entire vascular network and the higher differentially expressed genes also showed the distinct gene staining of arteries by anti-Sca-1, suggesting that AECs may be expression profile of AECs (both Nestin-GFPbright AECs and

2 NATURE COMMUNICATIONS | (2018) 9:2449 | DOI: 10.1038/s41467-018-04726-3 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04726-3 ARTICLE

Nestin-GFPdim AECs) compared to SECs, while Nestin-GFPbright of arterial-associated genes, including Efnb2, Gja5, Gja4, and Bmx AECs and Nestin-GFPdim AECs showed few genes that were than SECs (Fig. 2d). The higher expression of Bmx and Efnb2 in differentially expressed (Fig. 2c). Comparison of the expression AECs compared to SECs was confirmed independently using profile of AECs and SECs with perivascular stromal cells (labelled qPCR analysis (Fig. 2e, f). On the other hand, SECs were highly + by Lepr-Cre, NG2-Cre, or Myh11-CreER) or i.v.-stained CD31 enriched for the expression of the liver SECs genes Stab2, Dna- cells10, revealed separate clustering of AECs, SECs, and stromal se1l3, and C1qtnf121, as well as adhesion molecules mediating + cell types while total CD31 (i.v. injected antibody), as expected, haematopoietic stem and progenitor cell homing22, Sele, Selp, were between AECs and SECs (Supplementary Fig. 2). Impor- Icam1, Vcam1 compared to AECs (Fig. 2g). These data validate tantly, RNA-seq confirmed that Sca-1 (encoded by Ly6a) was the identity of AECs and SECs, and uncover their precise gene expressed by both AECs and SECs, and that its expression was signature (Fig. 2d, g). considerably higher in AECs (RPKMs in Nestin-GFPbright AECs, Surprisingly, we found that AECs expressed much higher levels Nestin-GFPdim AECs, and SECs: 2455 ± 1302, 2163 ± 780, and of the major HSC niche factors Scf (encoded by Kitl) and Cxcl12 362 ± 197, respectively), which is consistent with our imaging and (Fig. 2h and Supplementary Fig. 3). This was confirmed by qPCR FACS analysis results. AECs were highly enriched for expression analysis of sorted AECs and SECs (Fig. 2i, j). To obtain insight

ab 30

20 0.29 99.40 # cells 10 CD45/Ter-119 0.10 Sca-1 CD31/VE-cadherin Merge 0 CD31/VE-cadherin (i.v) Sca-1 c 3.8

77.0 Sca-1 Sca-1 CD31/VE-cadherin Merge

CD45/Ter-119 0.13 11.0 e

) 9.3 AECs

d CD31 PDPN (i.v) i.v 99.4 SECs Sca-1 85.9 CD45/Ter-119

0.12 VE-cadherin (

Isolectin (i.v) SSA PDPN (i.v) CD31/VE-cadherin PDPN 8.1 AECs ) i.v 99.8 SECs Sca-1 85.6 Isolectin (

CD45/Ter-119 0.12 Nestin-GFP Merge VE-cadherin (i.v) SSA PDPN (i.v)

f AECs SECs DN AECs 95.3 100 0.7 8.4

SECs 61.4 Sca-1 21.9 CD45/Ter-119 DN VE-cadherin (i.v)

CD31 PDPN (i.v) Isolectin (i.v)

Fig. 1 Separation of arterial from sinusoidal bone marrow endothelial cells using PDPN and Sca-1 expression. a Representative FACS plot of the Sca-1 expression on ECs from mice injected i.v. with anti-CD31/anti-VE-cadherin showing that all ECs are Sca-1+. Cells were pre-gated on singlet, live cells. b Representative whole-mount image of sternum from mice treated as in (a). Scale bar, 10 μm. c PDPN and Sca-1 separate CD45− Ter119− CD31+ cells into three populations: Sca-1high PDPN−, PDPN+ Sca1dim, and Sca-1− PDPN− double-negative populations. Cells were pre-gated on singlet, live cells. d Representative imaging of femur BM from Nestin-GFP mice stained with anti-VE-cadherin and anti-PDPN by i.v. injection showing that PDPN labels exclusively Nestin-GFP− sinusoids. Scale bar, 100 μm. e Representative FACS plot of ECs from a wild-type mouse treated with ﬂuorescently-labelled anti- VE-cadherin and Isolectin GS-IB4 i.v. to label ECs. Virtually all the ECs are within the AECs and SECs compartments. Cells were pre-gated on singlet, live cells. f Representative FACS plot of ECs from the same animals as in (e). All cells within the AECs and SECs gates are VE-cadherin+ Isolectin GS-IB4+ ECs. Double-negative cells are VE-cadherin− Isolectin GS-IB4−. Cells were pre-gated on singlet, live cells

NATURE COMMUNICATIONS | (2018) 9:2449 | DOI: 10.1038/s41467-018-04726-3 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04726-3 into their expression at the single cell level, we analysed ECs in higher levels in AECs, compared to SECs (Fig. 2h). These results Scf-GFP and Cxcl12-GFP reporter mice7,23. In these genetically suggest that AECs may play a more significant role than SECs in engineered animals, the GFP sequence was knocked into the HSC maintenance. niche gene locus to reflect the promoter activity. Remarkably, the Consistent with the higher expression of adhesion molecules + vast majority of AECs were Scf-GFP whereas virtually all SECs on SECs revealed by RNAseq analysis, the higher expression of − were Scf-GFP (Fig. 2k). Additionally, Cxcl12-GFP was expressed Sele, Selp, Icam1, Vcam1 in SECs compared to AECs was also at higher level in AECs than SECs (Fig. 2l), and several other confirmed using qPCR (Fig. 3a), leading us to evaluate whether known niche factors (e.g. Jag1, Jag2, Tgfb2, Il6) were expressed at they could substitute for PDPN. Indeed, we found that antibodies

abcAEC SEC

bright 100 Nestin-GFP Nestin-GFPbright AECs Nestin-GFPdim SECs 0 dim 57.7 34.8 97.6 0.1 Nestin-GFP

FSC 2 AECs

PC2 (24.0%) –100 SECs 0

GFP –200 0 200 –2 PC1 (32.4%) d gh AECs enriched genes SECs enriched genes Niche genes

Efnb2 Stab2 Kitl Bmx 2 Dnase1l3 2 Cxcl12 2 Gja5 C1qtnf1 Jag1 Gja4 0 Selp 0 Jag2 0 Icam1 Angpt1 Cd200 –2 –2 Cxcl12 Sele –2 ll7 Kitl Vcam1 Tgfb1 Gkn3 Epor Tgfb2 Fbln2 Cfd Tgfb3 Stmn2 Rarres2 ll6 Gper1 Aldob Sele Tmem255b Rgs4 SECs Nestin-GFPbright Nestin-GFPdim C1qtnf9 Adamts5 AECs AECs Timp4 Ackr1 1.5 Tmem100 Crygn ef0.4 *** Sparcl1 Prelp *** )

Slc26a10 Clca1 ) Rbp7 Gpm6a 0.3 Alpl Ubd 1.0 Aqp7 Mrc1 Gapdh

Hey1 Tgfbi Gapdh Emp2 Abcc9 0.2 mRNA Lims2 Serpina3f mRNA Cav1 Ctgf 0.5 Efnb2 Ednrb Angptl4 Bmx Dusp26 Sirpa 0.1 (relative to Ace Ap1b1 (relative to Car7 Serpina3g Pcp4l1 Gpr182 0 0 Edn1 Flt4 SECs AECs Sema3g Tspan7 SECs AECs Efhd1 Pparg Jam2 Kcnj8 4 Cd34 Cd302 ij1.0 Ramp3 Slc16a1 **** ) ***

Plat Slc31a1 ) Ltbp4 0.8 3 Ackr3

Col18a1 0.6 Gapdh Gapdh Serpinf1 mRNA 2 Acta2 Podxl mRNA 0.4 Gprc5a Scf Lmcd1 Cxcl12 1

Vegfc 0.2 (relative to Nxpe4 (relative to Igfbp3 0 0 SECs AECs SECs AECs kl Wild type AECs Wild type AECs

Cxcl12-GFP Scf-GFP AECs AECs

Wild type SECs Wild type SECs

SECs SECs Scf-GFP Cxcl12-GFP

0 2 3 4 5 10 10 10 10 Cxcl12-GFP Scf-GFP

4 NATURE COMMUNICATIONS | (2018) 9:2449 | DOI: 10.1038/s41467-018-04726-3 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04726-3 ARTICLE against either ICAM-1 or E-selectin (CD62E) marked all SECs by multipotent progenitor (LMPP), common myeloid progenitors FACS analysis (Fig. 3b). In addition, anti-ICAM-1 or anti-CD62E (CMP), common lymphoid progenitor (CLP), and granulocyte- antibodies along with anti-Sca-1 could be used to stain all ECs macrophage progenitor (GMP) (Supplementary Fig. 6b and c). without the need for intravenous antibody injection (Fig. 3c). These data, together with the expression results, clearly show that Furthermore, we have confirmed the high expression of adhesion the SCF derived from SECs is dispensable for HSC maintenance. molecules P-selectin, E-selectin, and ICAM-1 on SECs by fl – immuno uorescence microscopy analysis (Fig. 3d f). Thus, these AEC-derived SCF regulates HSC maintenance.AsBmx,a fi results indicate that BM AECs and SECs can be identi ed or member of the Tec tyrosine kinase family27, was selectively isolated at high purity by ex vivo staining using only four expressed by BM AECs in our RNAseq and qPCR analyses antibodies against CD45, Ter-119, Sca-1, and ICAM-1 or E- (Fig. 2d, e), we crossed Bmx-CreERT2 transgenic mice28 selectin. with the iTdtomato reporter and Nestin-GFP mice to evaluate whether Bmx-CreERT2 can target arteries. Femoral BM immuno- fl SEC-derived SCF is dispensable for HSC maintenance. Among uorescence microscopy imaging of triple-transgenic mice ERT2 bright the differentially expressed genes, we noted that the ery- revealed that Bmx-Cre exclusively labelled Nestin-GFP thropoietin receptor (Epor) was selectively expressed on SECs arteries, but not SECs (Fig. 5a, b; Supplementary Fig. 7). We also ERT2 (Fig. 2g). Although Epor expression is well described on erythroid injected anti-Sca-1 i.v. into Bmx-Cre ; iTdtomato mice to ERT2 lineage cells24 and was recently shown to be expressed by BM stain AECs and again found that the Bmx-Cre exclusively bright ECs25, its specific expression on SECs has, to our knowledge, not labelled Sca-1 arteries (Fig. 5c). FACS analysis of BM extracts been previously reported. qPCR analyses from sorted AECs and of these mice revealed that 62.7 ± 0.7% of AECs were targeted by fi fi Bmx-CreERT2, whereas SECs were not labelled (0.01 ± 0.03%; SECs con rmed that SECs expressed signi cantly higher Epor − − − + gene transcripts than AECs (Supplementary Fig. 4a). Imaging of Fig. 5d, e). In addition, CD45 Ter119 CD31 Nestin-GFP ERT2 femoral BM from Epor-Cre transgenic mice26 intercrossed with MSCs were not targeted by Bmx-Cre (Fig. 5f). Thus, Bmx- ERT2 fi ROSA26-loxP-stop-loxP-tdTomato (iTdTomato) reporter and Cre can speci cally label BM AECs, but not SECs or MSCs. Nestin-GFP mice (to mark arteries) revealed that SECs were To investigate the contribution of AECs in SCF secretion, we ERT2 fl/− targeted by Epor-Cre, while the arterial tree was mostly spared bred Bmx-Cre mice with Scf animals. We found that Scf ERT2 fi (Fig. 4a, b and Supplementary Fig. 4b). Among BM haemato- deletion using Bmx-Cre signi cantly reduced Scf expression poietic cells, labelling by Epor-Cre was largely restricted to in AECs, while it did not alter Scf expression in MSCs megakaryocyte-erythroid progenitors (MEP; Supplementary (Supplementary Fig. 8a). Arterial Scf deletion also did not alter Fig. 5), but these haematopoietic progenitors and their progeny EC numbers or the vascular structure of the BM (Supplementary were not visible in our immunofluorescence analyses due to the Fig. 8b and c). FACS analysis of BM mononuclear cells showed + − fi short exposure imposed by the strong iTdTomato signal in ECs. that CD150 CD48 LSK HSCs were signi cantly reduced both ERT2 FACS analysis of ECs from these mice showed that Epor-Cre in percentage and absolute numbers in the BM of Bmx-Cre + fl/− ERT2 − fl/− labelled 80.0 ± 2.1% of SECs and a minor proportion of AECs ( ) Scf mice compared to Bmx-Cre ( ) Scf littermate (21.0 ± 1.1%; Fig. 4c, d). Importantly, Epor-Cre did not target controls (Fig. 5g, h). LSK cells were similarly reduced (Fig. 5i, j). − − − + fi CD45 Ter119 CD31 Nestin-GFP MSCs (Fig. 4e). Thus, Competitive repopulation assay revealed a signi cant tri-lineage ERT2 + Epor-Cre selectively targets SECs in the BM microenvironment. reduction in reconstitution of BM cells from Bmx-Cre ( ) fl/− ERT2 − To assess the contribution of SEC-derived SCF in the Scf mice compared to cells from control Bmx-Cre ( ) Scf fl/− – fi maintenance of HSCs, we intercrossed Epor-Cre mice with Scf animals (Fig. 5k n). There was no signi cant difference in the fl − / mice and evaluated its impact on the number and function of total BM cellularity, numbers of MEP, MPP, LMPP, GMP, CMP, ERT2 + fl/− HSCs and progenitors. Deletion of Scf using Epor-Cre did not and CLP in the Bmx-Cre ( ) Scf mice compared to + + alter Scf expression by AECs or MSCs (MSCs: CD51 PDGFRα littermate controls (Supplementary Fig. 8d and e). Furthermore, − − ERT2 CD31 Ter119 ; Supplementary Fig. 6a). We found no Scf deletion by Bmx-Cre did not change the cell cycle status significant difference in the percentage or absolute numbers of of HSCs or the HSC localisation relative to arterioles (Supple- + − CD150 CD48 LSK HSCs in the BM between Epor-Cre (−) Scf mentary Fig. 8f and g). These data, together with the expression fl − fl − fi / and Epor-Cre (+) Scf / littermates (Fig. 4f, g). Competitive pro le, clearly indicate that BM EC SCF is derived from the repopulation experiments also showed that equal repopulation arterial vascular tree. fl − capacity of BM cells obtained from Epor-Cre (−) Scf / and fl − Epor-Cre (+) Scf / mice (Fig. 3h–k). In addition, we did not AECs self-regenerate and do not regenerate SECs. The origin of observe any significant difference in total BM cellularity or ECs during the BM regeneration remains unclear. For example, − + + progenitor counts, including Lin , Sca-1 , c-kit cells (LSK), following irradiation SECs are heavily damaged whereas arteries multipotent progenitor cells (MPP), MEP, lymphoid-primed are thought to be more structurally resilient. Type H ECs were

Fig. 2 AECs, but not SECs, express high levels of canonical niche factors. a Representative FACS plot of AECs and SECs from Nestin-GFP mice. SECs, Nestin- GFPbright AECs and Nestin-GFPdim AECs were sorted for RNA sequencing. b Principal component analysis of gene expression, showing that Nestin-GFPbright AECs and Nestin-GFPdim AECs cluster together, while SECs occupy a space different from AECs. c Hierarchical clustering of significantly differentially expressed genes. The values shown are z-scores. d Heatmap of the genes specific in AECs. Groups are indicated by the colour of the bar on top of the graph as used in panels (a), (b), and (c). The values are log10 transformed-RPKM. e qPCR was performed on sorted AECs and SECs to quantify mRNA levels of Bmx. n = 7 for AECs and SECs. f qPCR was performed on sorted AECs and SECs to quantify mRNA levels of Efnb2. n = 7 for AECs and SECs. g Heatmap of genes specific in SECs. Groups are indicated by the colour of the bar on top of the graph as used in panel c. The values are log10 transformed- RPKM. h Heatmap of niche genes. Different groups are indicated by the colour of the bar on top of the graph as used in panels (a), (b), and (c). The values are log10 transformed-RPKM. i qPCR was performed on sorted AECs and SECs to quantify mRNA levels of Scf. n = 11 for AECs and SECs. j qPCR was performed on sorted AECs and SECs to quantify mRNA levels of Cxcl12. n = 12 for AECs and SECs. k Representative FACS plot of Scf-GFP reporter activity in AECs and SECs. l Representative FACS plot of Cxcl12-GFP reporter activity in AECs and SECs. e, f, i, j Data are represented as mean ± SEM. Data were analysed with two-tailed t-test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

NATURE COMMUNICATIONS | (2018) 9:2449 | DOI: 10.1038/s41467-018-04726-3 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04726-3

ab3 Overlay 0.4 *** **** ) ) AECs 0.3 2 Gapdh Gapdh mRNA mRNA 0.2 SECs Sca-1 Sca-1 1 Icam1 Vcam1 AECs 0.1 (relative to (relative to SECs 0 0 SECs AECs SECs AECs PDPN (in vivo) ICAM-1

0.4 **** 2.0 **** AECs ) ) 0.3 1.5 Gapdh Gapdh

0.2 1.0 Sca-1 Sca-1 mRNA mRNA AECs Sele Selp 0.1 0.5 SECs

(relative to SECs (relative to

0 0 ICAM-1 E-selectin SECs AECs SECs AECs

cdAECs SECs CD31/CD144 Nestin-GFP 0.13 SECs AECs Sca-1 1.1 92.4 95.0

ICAM-1 0.07 P-selectin Merge CD45/Ter-119

AECs SECs Sca-1 0.55 93.6 95.2

E-selectin CD31 ef CD31/CD144 Nestin-GFP CD31/CD144 Sca-1

ICAM-1 Merge E-selectin Merge

Fig. 3 Elevated cell adhesion molecule expression on SECs. a qPCR analysis of adhesion molecules, Selp, Sele, Icam1, and Vcam1 of sorted AECs and SECs. Data are represented as mean ± SEM. n = 4 for AECs and SECs. Data were analysed with two-tailed t-test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. b PDPN can be substituted for antibody against the adhesion molecule ICAM1 and E-selectin. Shown is EC isolation from a wild-type mouse injected i.v. with anti-PDPN and the enzymatically dissociated bone marrow cells were stained with other antibodies ex vivo and pre-gated on singlet, live, CD45−, Ter119−, CD31+ cells. c Anti-CD31 is dispensable for the analysis of bone marrow AECs and SECs. Flushed enzymatically dissociated bone marrow cells were stained with anti-CD45, anti-Ter119, anti-ICAM-1/anti-CD62E, anti-CD31, and anti-Sca-1 ex vivo. Analysis was pre-gated on singlet, live, CD45−, Ter119− cells. d P-selectin marks sinusoids. Anti-P-selectin and anti-CD31/CD144 were i.v. injected into Nestin-GFP mice and femur was imaged after removal. Scale bar, 50 um. e ICAM-1 marks sinusoids. Anti-ICAM-1 and anti-CD31/CD144 were i.v. injected into Nestin-GFP mice and femur was imaged after harvest. Scale bar, 100 um. f E-selectin marks sinusoids. Anti-E-selectin, anti-Sca-1, and anti-CD31/CD144 were i.v. injected into wild-type mice and femur was imaged after harvest. Scale bar, 50 um

6 NATURE COMMUNICATIONS | (2018) 9:2449 | DOI: 10.1038/s41467-018-04726-3 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04726-3 ARTICLE

a b

Epor -Cre;iTdTomato

Bone Epor-Cre;iTdTomato Nestin-GFP

Nestin-GFP

CD31/VE-cadherin Merge CD31/VE-cadherin – – + c CD45 Ter119 CD31 AECs SECs d 100 AEC 81 80 21 60

SSA 40 Sca1 Merge SEC (%) -Cre/iTdTomato 20

Epor 0 PDPN (i.v.) Epor-Cre;iTdTomato AECs SECs fl/– fl/– e Epor-Cre (–) Scf Epor-Cre (+) Scf fg0.03 31.3 4

0.02 3 /femur) 3 SSA SSA 2 10 × CD45/Ter-119 0.43 1.1 HSC (%) 0.01 1

CD31 Nestin-GFP Epor-Cre;iTdTomato HSC ( 0 0

Epor-Cre (–) Scf fl/– Epor-Cre (+) Scf fl/–

hi80 80 j80 k50 40 60 60 60 30 40 40 40 20 cells (%) T cells (%) 20 20 B cells (%) 20 10 CD45.2 derived CD45.2 derived CD45.2 derived CD45.2 derived myeloid cells (%) 0 0 0 0 4 8 12 16 4 8 12 16 4 8 12 16 4 8 12 16 Weeks after BMT Weeks after BMT Weeks after BMT Weeks after BMT

Fig. 4 SCF secreted by SECs is dispensable for HSC maintenance. a Images of femur from Epor-Cre;iTdtomato;Nestin-GFP mice stained i.v. with anti-VE- cadherin and anti-CD31. All the panels show the same field for different channels. Scale bar, 100 μm. b Enlarged images of the field highlighted by yellow rectangles in (a). All panels show the same field for different channels. The Epor-Cre-labelled cells do not colocalise with Nestin-GFPbright arteries. Scale bar, 10 μm. c Representative FACS plot of the labelling of AECs and SECs by Epor-Cre. d Quantification of the labelling efficiency of AECs and SECs by Epor-Cre. Data are represented as mean ± SEM. n = 7 for AECs and SECs. e Representative FACS plot of Nestin-GFP+ MSCs labelling by Epor-Cre. f Percentage of HSCs (CD150+ CD48− LSK) in the BM. Data are represented as mean ± SEM. n = 4 for Cre (−) and n = 7 for Cre (+). g Number of HSCs (CD150+ CD48 − LSK) in the BM. Data are represented as mean ± SEM. n = 4 for Cre (−) and n = 7 for Cre (+). h–k Percentage of donor-derived total cells and tri-lineage (myeloid, B and T) cells after competitive transplantation. n = 7 for Cre (−), and n = 3 for Cre (+). The label of X-axis indicates the time (weeks) after transplantation reported to give rise to other ECs during the juvenile stage, but we treated mice with fluorouracil (5-FU) and monitored endo- these results were based on the differential labelling of ECs in thelial recovery (Fig. 6a). 5-FU treatment significantly reduced the Cdh5-CreER mice13. We thus evaluated whether regeneration of number of SECs and AECs (Fig. 6b). Both AECs and SECs SECs was dependent on arterial-derived endothelial precursors. robustly proliferated on day 9 after 5-FU challenge, whereas by Bmx-CreERT2;iTdtomato mice were treated with tamoxifen contrast, most ECs were quiescent in the steady state (Fig. 6c). 4 weeks prior to myeloablation28. In the first set of experiments, FACS analysis of BM ECs revealed that Bmx-labelled ECs

NATURE COMMUNICATIONS | (2018) 9:2449 | DOI: 10.1038/s41467-018-04726-3 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04726-3

a b

CD31/VE-cadherin Nestin-GFP CD31/VE-cadherin Nestin-GFP

Bmx-CreERT2;iTdtomato Merge

Bmx-CreERT2;iTdtomato Merge c

d CD45–Ter119–CD31+ AECs SECs

18.1 68.0 0 2.9 CD31/VE-cadherin Sca-1 AEC Sca1

SEC SSA

Bmx-CreERT2;iTdtomato Merge PDPN (i.v.) Bmx-CreERT2;iTdtomato e 80 f 60 0.8

α 12.3 ;iTdtomato 40 SSA ERT2 PDGFR -Cre

20 CD45/Ter-119 1.0 Bmx

0 CD51 AEC SEC CD31 Bmx-CreERT2;iTdtomato

Bmx-CreERT2(–) Scf fl/– Bmx-CreERT2(+) Scf fl/–

ghij0.05 ** 4 **0.4 **2.5 *** 0.04 2.0 3 0.3

/femur) 1.5 0.03 /femur) 4 3 2 0.2 10 10 × 0.02 × 1.0 LSK (%) HSC (%) 1 0.1 0.5

0.01 LSK ( HSC ( 0 0 0 0 kl mn 80 80 ** 80 *** * * ** 80 ** ** ** ** 60 60 60 60 40 40 40 40 *

20 B cells (%) 20 20 T cells (%) 20 CD45.2 derived CD45.2 derived myeloid cells (%) CD45.2 derived 0 0 CD45.2 derived cells (%) 0 0 4 8 12 16 4 8 12 16 4 8 12 16 4 8 12 16 Weeks after BMT Weeks after BMT Weeks after BMT Weeks after BMT

Fig. 5 SCF from AECs regulates HSCs. a Representative images of femur from Bmx-CreERT2;iTdtomato;Nestin-GFP mice stained with anti-VE-cadherin and anti-CD31. All the panels show the same field for different channels. Scale bar, 100 μm. b Enlarged image of the field highlighted by the yellow rectangle in (a). All panels show the same field for different channels. The Bmx-CreERT2-labelled cells colocalise with Nestin-GFP-labelled cells. Scale bar, 10 μm. c Image of whole-mount sternal BM from Bmx-CreERT2;iTdtomato injected with anti-Sca-1. Scale bar, 50 μm. d Representative FACS plot of the labelling of AECs and SECs by Bmx-CreERT2. e Quantification of the labelling efficiency of AECs and SECs by Bmx-CreERT2. Data are represented as mean ± SEM. n = 8 for AECs and SECs. f Representative FACS plot of Nestin-GFP+ MSCs labelling by Bmx-CreERT2. g Percentage of HSCs (CD150+ CD48− LSK) in the BM. Data are represented as mean ± SEM. n = 11 for Cre (−) and n = 14 for Cre (+). h Number of HSCs (CD150+ CD48− LSK) in the BM. Data are represented as mean ± SEM. n = 11 for Cre (−) and n = 14 for Cre (+). i Percentage of LSK in the BM. Data are represented as mean ± SEM. n = 11 for Cre (−) and n = 14 for Cre (+). j Number of LSK in the BM. Data are represented as mean ± SEM. n = 11 for Cre (−) and n = 14 for Cre (+). k–n Percentage of donor-derived cells after competitive transplantation. n = 6 mice for Cre (−), and n = 7 for Cre (+). The label of X-axis indicates the time (weeks) after transplantation. g–n Data were analysed with two-tailed t-test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

8 NATURE COMMUNICATIONS | (2018) 9:2449 | DOI: 10.1038/s41467-018-04726-3 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04726-3 ARTICLE

ab40 e

ERT2 Bmx-Cre ;iTdtomato: Day 17 –4 weeks D0 D3 D9 D17 20 Tamoxifen 5-FU

FACS & imaging Relative to control CD31/VE-cadherinBmx-CreERT2;iTdtomato Merge analysis 0

AECs SECs cdAECs SECs Day 3 Day 9 Day 17 7.7 0 4.8 4.8

Control AECs 66.7% 52.3 66.7%

84.6 90.5 SSA Ki-67 43.4 18.2 68.0 8.3

5-FU SECs 0 0.62 0

38.6 23.4

Hoechst 3334 Bmx-CreERT2;iTdtomato g 1 month 4 month f Bmx-CreERT2;iTdtomato: –4 weeks Day 0 1 m 2 m 4 m 13 m

Tamoxifen Lethal irradiation and bone marrow ERT2 ERT2 transplantation FACS & imaging Bmx-Cre ;iTdtomato Bmx-Cre ;iTdtomato analysis CD31/VE-cadherin CD31/VE-cadherin

i 1 month 2 month 4 month h 42.9 40.8 21.7 Bmx-CreERT2;iTdtomato AECs 13 month SSA

0 0 0.7 CD31/VE-cadherin

SECs

ERT2 Bmx-Cre ;iTdtomato Merge

Fig. 6 AECs self-regenerate, and do not regenerate SECs. a Scheme of experiment design. Bmx-CreERT2;iTdtomato mice were injected with Tamoxifen to activate Cre expression. Four weeks later, 5-FU was given to these mice and BM ECs were analysed by FACS and immunofluorescence analysis. b Numbers of AECs and SECs 3 days after 5-FU treatment. Data are represented as mean ± SEM. c Representative FACS plot of cell cycle analysis of AECs and SECs from control mice and mice treated with 5-FU (9 days after 5-FU) using Hoechst 3334 and Ki-67. d FACS plot of the labelling of AECs and SECs by Bmx- CreERT2 at different time points after 5-FU treatment. e Representative image of whole-mount sternum from mice treated as in (a). The bone was harvested on day 17 after 5-FU injection. Mice were injected i.v. with anti-VE-cadherin and anti-CD31. All panels show the same area for different channels. Scale bar, 50 μm. f Scheme of experiment design. Bmx-CreERT2;iTdtomato mice were injected with Tamoxifen to activate Cre expression. Four weeks later, these mice were lethally irradiated and transplanted with BM cells from wild-type mice. BM ECs were analysed by FACS and immunofluorescence analysis at different time points after lethal irradiation. g Representative images from whole-mount sternum at 1 month (left panel) and 4 months (right panel) after lethal irradiation. Mice were injected i.v. with anti-VE-cadherin and anti-CD31. Scale bar, 200 μm. h Images from whole-mount sternum in which ECs are stained i.v. by anti-VE-cadherin and anti-CD31 injection at 13 months after lethal irradiation. The panels on the right show the same field for different channels. Scale bar, 100 μm in the left panel and 10 μm in the right panel. i FACS plot of the labelling of AECs and SECs by Bmx-CreERT2 at different time points after lethal irradiation

NATURE COMMUNICATIONS | (2018) 9:2449 | DOI: 10.1038/s41467-018-04726-3 | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04726-3

– remained exclusively in the AECs compartment during regen- also represent a specific niche promoting HSC quiescence33 35. eration (Fig. 6d). Immunofluorescence analyses also revealed that Others have argued that quiescent HSCs are uniformly dis- + iTdtomato cells were confined to AECs 17 days after 5-FU tributed throughout the BM near sinusoids and that arteriole- challenge (Fig. 6e). These data suggests that AECs and SECs may associated stromal cells did not play a specific role in HSC regenerate separately after 5-FU-induced regeneration. maintenance36. The selective expression of SCF and other niche To study the vascular regeneration over a longer timeframe, factors by AECs clearly support further the notion that arterioles we lethally irradiated tamoxifen-treated Bmx-CreERT2;iTdto- confer a specific microenvironment for distinct HSC subsets. This mato mice and transplanted them with wild-type BM cells for idea is also consistent with recent studies showing that lineage- haematopoietic rescue (Fig. 6f). Animals were evaluated at biased HSCs were located in distinct niches in which HSCs various time points after a lethal dose of irradiation. Similar to expressing von Willebrand factor (vWF) were associated with and − the 5-FU model, we found that labelled ECs remained within regulated by megakaryocytes, whereas vWF HSCs were found to the AECs compartment and did not contribute to SECs be associated with and regulated by arteriolar perivascular cells37. regeneration even 13 months post-irradiation (Fig. 6g, h). It will be important in future investigations to address the FACS analysis of BM ECs also showed that the Bmx-CreERT2- implications of these distinct niches in the development and labelled cells only gave rise to AECs (Fig. 6i). Taken together, progression of haematological diseases. these results indicate that AECs and SECs self-regenerate independently from specified radio-resistant endothelial pre- Methods cursors in the adult BM. Mice. B6.Cg-Gt(ROSA)26Sortm14(CAG–tdTomato)Hze/J (iTdTomato) and Kitltm1.1Sjm/J (Scf-GFP) mice were purchased from Jackson Laboratory, Cxcl12- fl fl GFP mice and Cxcl12 / mice were a gift from T. Nagasawa23, and fl fl AEC-derived SCF regulates HSC regeneration. To investigate Scf / mice were a gift from S. Morrison7. C57BL/6J-Tg(Bmx-CreERT2)1Rha mice 28 whether AECs-derived SCF contributes to HSC regeneration after were kindly provided by Dr. Ralf Adams . Epor-merCre mice were obtained from fl − 26 38 myeloablation, we treated the Bmx-CreERT2 (+) Scf / mice and Dr. Ann Mullally . Nestin-GFP mice are bred in our facilities . C57BL/6-CD45.1/ ERT2 fl/− 2 congenic strains were purchased from Jackson Laboratory. All mice were Bmx-Cre (−) Scf with 5-FU and analysed HSCs recovery maintained in pathogen-free conditions under a 12 h:12 h light/dark cycle and fed at different time points after treatment (Fig. 7a). We found that ad libitum. All experiments were carried out using gender-matched littermate HSC regeneration was significantly impaired in Bmx-CreERT2 (+) controls. All mice were analysed at 6–14 weeks of age. Both genders were used for fl/− ERT2 − fl/− experiments. All experimental procedures were approved by the Animal Care and Scf mice compared with Bmx-Cre ( ) Scf at day 21 Use Committees of Albert Einstein College of Medicine. after 5-FU treatment (Fig. 7b). Furthermore, reconstitution ERT2 + fl/− capacity of day 21 HSCs from Bmx-Cre ( ) Scf mice was Tamoxifen treatment. For induction of Bmx-CreERT2-mediated recombination, significantly reduced compared to HSCs from Bmx-CreERT2 (−) fl − we adopted the methods used by the lab generating this mice to target arteries with Scf / (Fig. 7c). Thus, AEC-derived SCF contributes to HSC modification28. Briefly, 2-week-old mice were injected with tamoxifen (Sigma) − regeneration after 5-FU treatment. intraperitoneally at 100 mg kg 1 dissolved in corn oil (Sigma) daily for two rounds of 5 consecutive days with a 7-day interval.

Discussion Competitive repopulation assays. Competitive transplantation was performed Recent studies have suggested that multiple stromal and haema- using the CD45.1/CD45.2 congenic system. Same number of BM cells (3 × 105) topoietic cells contribute to HSC maintenance. Although HSCs collected from gene-deleted mice or control mice (CD45.2) were transplanted into lethally irradiated (6Gy + 6Gy) CD45.1 recipients with 3 × 105 competitor are broadly distributed close to sinusoids, the exact role of SECs CD45.1 cells. Peripheral blood collection was carried out by retro-orbital plexus in HSC maintenance has not been specifically demonstrated. bleeding with a heparinised microcapillary tube under deep anaesthesia to 0.5 M Cultured AKT-activated ECs (which are likely a mixture of AECs EDTA solution. CD45.1/CD45.2 chimerism of recipients’ blood was analysed for and SECs) can expand HSPCs both in vitro and in vivo29,30. up to 4 months after transplantation. Specific deletion of the Notch1 ligand Jag1 in all ECs using constitutive Cdh5-Cre reduced BM HSC numbers in steady state Flow cytometry and cell sorting. The processing of BM cells were previously reported10. Briefly, for the analyses of haematopoietic cells, BM cells were flushed and impaired haematopoietic recovery after sublethal irradia- and dissociated by gently passing through a 21G needle. For analysis of stromal 31 − tion . Deletion of Scf in the both the endothelial (Tie2-Cre) and cells and ECs, BM cells were flushed and digested with 1 mg ml 1 collagenase IV perivascular stromal (Lepr-Cre) compartments have suggested (Gibco), and 2 mg/ml dispase (Gibco) in Hank’s balanced salt solution (Gibco) for 40 min at 37 °C. Ammonium chloride was used for red blood cell (RBC) lysis. Then additive contributions of these niches in that the reduction of fi μ HSC numbers in steady-state BM was greater when Scf was cells were ltered through 100 m nylon mesh. For FACS analysis, cells were 7 stained with antibodies in PEB (PBS containing 0.5% BSA and 2 mM EDTA) buffer deleted in both compartments compared to either alone . Owing for 30 min at 4 °C. The following antibodies were used: allophycocyanin (APC)- to the similar spatial distribution of CAR cells and SECs in the anti-Gr-1 (eBioscience, Cat# 17-5931, clone RB6-8C5), phycoerythrin (PE)-anti- BM, it has been assumed that SECs might be the most important CD11b (eBioscience, Cat# 12-0112-83 clone M1/70), APC-eFluor780-anti-CD45R EC type for HSC maintenance. Our results show an unanticipated (eBioscience, Cat# 47-0452, clone RA3-6B2), PerCP-Cyanine5.5-anti-CD3e (eBioscience, Cat# 45-0031, clone 145-2C11), biotin-anti-Lineage (BD Biosciences, selective expression of niche factors in the arterial vascular tree, Cat# 559971, clones TER119, RB6-8C5, RA3-6B2, M1/70, 145-2C11), fluorescein supporting an important role for an arterial-associated HSC isothiocyanate (FITC)-anti-CD45.2 (eBioscience, Cat# 11-0454, clone 104), FITC- niche. anti-Ly6A/E (eBioscience, Cat# 11-5981, clone D7), PE-Cy7-anti-CD117 The presence of an arterial niche for HSCs has been con- (eBioscience, Cat# 25-1171, clone 2B8), Brilliant Violet 421-anti-mouse CD48 + − − − (Biolegend, Cat# 103427, clone HM48-1), Brilliant Violet 605-anti-mouse CD150 troversial. A subset of CD150 CD48 Lin CD41 HSCs was (Biolegend, Cat# 115927, clone TC15-12F12.2), PE-Cy7-anti-Ki67 (eBioscience, shown to significantly associate to arterioles compared to the Cat# 25-5698-80, clone SolA 15), PE-Cy7-anti-CD31 (eBioscience, Cat# 25-0311, random assignment of HSC-like dots11. Importantly, the pro- clone 390), APC-eFluor780-anti-CD45 (eBioscience, Cat# 47-0451, clone 30-F11), portion of arteriole-associated HSCs was altered during their APC-eFluor780-anti-TER119 (eBioscience, Cat# 47-5921, clone TER119), Alexa 11 Fluor 700-anti-CD16/32 (eBioscience, Cat# 56-0161-82, clone 93), PerCP- proliferation, suggesting a functional relevance . Differences in Cyanine5.5-anti-CD127 (eBioscience, Cat# 45-1271-80, clone A7R34), PE-anti- permeability of sinusoids and arteries have subsequently been CD135 (eBioscience, Cat# 12-1351, clone A2F10), streptavidin APC-eFluor780 proposed to regulate HSC proliferation whereby less permeable (eBioscience, Cat# 47-4317-82), Alexa Fluor 647-anti-CD144 (Biolegend, Cat# arterial vessels maintained an environment low in reactive oxygen 138006, clone BV13), Alexa Fluor 647-anti-CD31 (Biolegend, Cat# 102516, clone MEC13.3), PE-anti-mouse PDPN (Biolegend, Cat# 127408, clone 8.1.1), PE-Cy7- species (ROS) promoting HSC quiescence, while the more anti-mouse PDPN (Biolegend, Cat# 127412, clone 8.1.1), Alexa Fluor 488-anti- permeable sinusoidal vessels activated HSCs by high ROS due to mouse CD31 (Biolegend, Cat# 102514, clone MEC13.3), Pacific Blue-anti-mouse plasma leakage32. Further studies show that megakaryocytes may CD54 (Biolegend, Cat# 116116, clone YN1/1.7.4), PE-Rat Anti-Mouse CD62E (BD,

10 NATURE COMMUNICATIONS | (2018) 9:2449 | DOI: 10.1038/s41467-018-04726-3 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04726-3 ARTICLE

a Bmx-CreERT2 (–) Scf fl/– b Bmx-CreERT2 (+) Scf fl/– Lethally irradiated 4 * CD45.1mice –4 Weeks D0 D8 D12 D15 D21 3 D21 BM cells /femur)

(CD45.2) 3 Tamoxifen Peripheral blood 2 10 analysis × 5-FU FACS analysis 1 of HSPCs Competitor: BM cells HSC ( (CD45.1) 0 8 121521 cd e f 80 100 * 100 80 * * ** *** ** 80 *** ** 80 ** 60 ** *** *** 60 ** ** 60 60 40 40 40 40 * B cells (%) 20 T cells (%) 20 CD45.2 derived 20 20 CD45.2 derived myeloid cells (%) CD45.2 derived

CD45.2 derived cells (%) 0 0 0 0 4 81216 4 8 12 16 4 8 12 16 4 8 12 16 Weeks after BMT Weeks after BMT Weeks after BMT Weeks after BMT

Bmx-CreERT2 (–) Scf fl/– Bmx-CreERT2 (+) Scf fl/–

Fig. 7 SCF from AECs contributes to HSCs regeneration. a Scheme of experiment design. Bmx-CreERT2;Scf ﬂ/− mice and control mice were ﬁrstly given Tamoxifen to activate Cre expression. Four weeks later, 5-FU was administered and HSCs were analysed at different time points by FACS. b HSC number in femur at indicated time points. n = 12 mice for Cre (−), and n = 9 for Cre (+). c–f Percentage of donor-derived cells after competitive transplantation. n = 5 mice for Cre (−), and n = 7 for Cre (+). The label of X-axis indicates the time (weeks) after transplantation

Cat# 553751, clone 10E9.6), Alexa Fluor 700-Anti-mouse Ly-6A/E (Sca-1) Fixation/Permeabilization solution for 20 min, washed, and stained with anti-Ki-67 − (eBioscience, Cat# 56-5981-82, clone D7), PerCP-Cyanine5.5-anti-CD45 antibody and Hoechst 33342 (Sigma) at 20 μgml 1 in 1× BD Perm/Wash™ buffer (eBioscience, Cat# 45-0451-82, clone 30-F11), PerCP-Cyanine5.5-anti-Ter119 for 30 min. (eBioscience, Cat# 45-5921-82, clone TER119), mouse EphB4 Antibody (R&D, Cat# AF446), Donkey anti-Goat IgG (H+L) Secondary Antibody, Alexa Fluor 488 fl (Thermo Fisher Scientific, Cat#A-11055), and PE Syrian Hamster IgG Isotype Ctrl Whole-mount BM confocal immuno uorescence analyses of HSCs. Endogen- 11 fl Antibody (Biolegend, Cat# 402008, clone SHG-1). We used these antibodies at ous HSC staining in sternal BM was performed as previously described . Brie y, 1:100 dilution for FACS analysis. For analysis of EC by FACS, we injected i.v. 3 μg Alexa Fluor 647-anti-CD144 (Biolegend, cat# 138006, clone BV13, 1:100) and PE-anti-mouse PDPN or PE-Cy7-anti-mouse PDPN (Biolegend, Cat# 127412, Alexa Fluor 647-anti-CD31 (Biolegend, Cat# 102516, clone MEC13.3) were i.v. clone 8.1.1) to the mice and waited for 10 min before we sacrificed the mice for injected into mice to stain the vasculature. Sternal BM fragments were removed 10 fi FACS analysis. FACS analyses were carried out using an LSR II flow cytometer min after antibody injection, xed with 4% PFA for 30 min, rinsed with PBS for equipped with FACS Diva 6.1 software (BD Biosciences). Dead cells and debris three times, and then blocked/permeabilised in PBS containing 20% normal goat were excluded by FSC, SSC, and 4′, 6-diamino-2-phenylindole (DAPI) (Sigma) serum and 0.5% Triton X-100 at room temperature. Primary antibodies were staining. Cell sorting was performed on a FACSAria Cell Sorter (BD Biosciences). incubated for 2 overnights at room temperature. The primary antibodies used were Data were analysed with FlowJo (Tree Star) software. biotin-anti-lineage (TER119, RB6-8C5, RA3-6B2, M1/70, 145-2C11, BD Bios- ciences, cat. No. 559971, 1:100), biotin-anti-CD48 (eBioscience, cat# 13-0481, HM48-1, 1:100), biotin-anti-CD41 (eBioscience, cat# 13-0411, MWReg30, 1:3000), RNA isolation and quantitative PCR. Sorted cells were collected in Trizol (Sigma) PE-anti-CD150 (Biolegend, cat# 115904, clone TC15-12F12.2, 1:100). After rinsing and RNA was extracted following the manufacturer’s protocol with one mod- the tissue with PBS, the tissues were incubated with streptavidin eFluor 450 ification: linear acrylamide was added to the solution to facilitate RNA precipita- (eBioscience, cat# 48-4317, 1:100). tion. Reverse transcriptions were performed using RNA to cDNA EcoDry Premix (Clontech) according to the manufacturer’s protocols. Quantitative PCR was performed with SYBR Green (Roche) on the QuantStudio 6 Flex Real-time PCR RNA preparation and next-generation sequencing. Total RNA from sorted System (Applied Biosystems). The PCR protocol consisted of one cycle at 95 °C (10 Nestin-GFPbright AECs, Nestin-GFPdim AECs and SECs was extracted using the min) followed by 40 cycles of 95 °C (15 s) and 60 °C (1 min). Expression of Gapdh RNAeasy Plus Micro kit (Qiagen). The integrity and purity of total RNA were − was used as an internal control. The relative gene expression is determined as 2 (Ct assessed using an Aligent 2100 Bioanalyzer (Agilent Technologies). Com- − (gene) Ct(control)). Primers used include: Gapdh forward, 5′-CAC ATT GGG GGT plementary DNA was generated using the SMART-Seq v4 Ultra Low Input RNA AGG AAC AC-3′; Gapdh reverse, 5′-ACC CAG AAG ACT GTG GAT GG-3′; Kit for Sequencing (Clontech Laboratories) from total RNA. The Nextera XT DNA Cxcl12 forward, 5′-CGC CAA GG TCG TCG CCG-3′; Cxcl12 reverse, 5′-TTG Sample Preparation Kit (Illumina) was used for the preparation of DNA libraries. GCT CTG GCG ATG TGG C-3′; Scf forward, 5′-CCC TGA AGA CTC GGG CCT The libraries were then submitted for Illumina HiSeq2500 sequencing (Illumina) A-3′; Scf reverse, 5′-CAA TTA CAA GCG AAA TGA GAG CC-3′; Bmx forward, according to standard operating procedures. 5′-GCA AAA AGA GAT ACG GGG CAA-3′; Bmx reverse, 5′-GAA CTT TCC ATC CAC AAA GAA GC-3′; Efnb2 forward, 5′-ATT ATT TGC CCC AAA GTG GAC TC-3′; Efnb reverse, 5′-GCA GCG GGG TAT TCT CCT TC-3′; Epor for- RNA-Seq analysis. RNA-Seq results generated from Illumina HiSeq 2500 were ward, 5′-GGT GAG TCA CGA AAG TCA TGT-3′; Epor reverse, 5′-CGG CAC processed following a recently developed reproducible pipeline by Single Cell AAA ACT CGA TGT GTC-3′; Icam1 forward, 5′-GGA CCA CGG AGC CAA TTT Genomics and Epigenomics Core facility at Albert Einstein College of Medicine. ′ ′ ′ Briefly, single-end sequencing reads were aligned to the mouse genome using C-3 ; Icam1 reverse, 5 -CTC GGA GAC ATT AGA GAA CAA TGC-3 ; Vcam1 39 forward, 5′-GAC CTG TTC CAG CGA GGG TCT A-3′; Vcam1 reverse, 5′-CTT Spliced Transcripts Alignment to a Reference (STAR) . We next used htseq-count ′ ′ to assign aligned reads to genes. Reads per kilobase of transcript per million CCA TCC TCA TAG CAA TTA AGG TG-3 ; Selp forward, 5 -GGT ATC CGA fi 40 AAG ATC AAC AAT AAG TG-3′; Selp reverse, 5′-GTT ACT CTT GAT GTA (RPKM) was used as the expression quanti cation method . The RPKM matrix GAT CTC CAC ACA-3′; Sele forward, 5′-CCC TGC CCA CGG TAT CAG-3′; Sele was log2-transformed. reverse, 5′-CCC TTC CAC ACA GTC AAA CGT-3′. To identify genes that could be reliably used for analysis of each cell population, we first filtered out genes that had a counts per million < 0.1 in at least seven of the nine samples. 17,232 genes remained after filtering. Next, we normalised for RNA Cell cycle analyses. Cell cycle analyses were performed using BD Cytofix/Cyto- composition by calculating scaling factors for each sample using calcNormFactors perm™ Fixation/Permeabilization Kit following the manufacturer’s protocol. in edgeR (version 3.12.0). The default method for computing scaling factors in Briefly, BM cells were stained with surface markers, fixed and permeabilised in the calcNormFactors, trimmed mean of M-values (TMM) between each pair of

NATURE COMMUNICATIONS | (2018) 9:2449 | DOI: 10.1038/s41467-018-04726-3 | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04726-3 samples, was used. The original library size, before normalising, was calculated as 20. Kwak, H. et al. Sinusoidal ephrin receptor EPHB4 controls hematopoietic the sum of the counts in each sample. The effective library size of each sample is progenitor cell mobilization from bone marrow. J. Clin. Invest. 126, the product of the original library size and the scaling factor for a given sample. 4554–4568 (2016). Next, the estimate Disp function in edgeR was used to calculate the dispersions of 21. Nonaka, H., Sugano, S. & Miyajima, A. Serial analysis of gene expression in the negative binomial model using the quantile-adjusted conditional maximum sinusoidal endothelial cells from normal and injured mouse liver. Biochem. likelihood (qCML) method. Gene-wise exact tests were performed to test whether a Biophys. Res. Commun. 324,15–24 (2004). gene was differentially expressed between any two conditions. Nominal p-values 22. Frenette, P. S., Subbarao, S., Mazo, I. B., von Andrian, U. H. & Wagner, D. D. were corrected using the Benjamini–Hochberg procedure to adjust for multiple Endothelial selectins and vascular cell adhesion molecule-1 promote hypothesis testing. hematopoietic progenitor homing to bone marrow. Proc. Natl. Acad. Sci. U.S. A. 95, 14423–14428 (1998). Statistics and reproducibility. No statistical method was used to predetermine 23. Sugiyama, T., Kohara, H., Noda, M. & Nagasawa, T. Maintenance of the sample size. The experiments were not randomised and investigators were not hematopoietic stem cell pool by CXCL12-CXCR4 chemokine signaling in blinded to allocation during experiments and outcome analyses. All data are bone marrow stromal cell niches. Immunity 25, 977–988 (2006). represented as mean ± SEM. n represents mouse number analysed in each 24. Constantinescu, S. N., Ghaffari, S. & Lodish, H. F. The erythropoietin experiment, as detailed in figure legends. We used two-tailed Student’s t-tests for receptor: structure, activation and intracellular signal transduction. Trends fi evaluating the signi cance of difference unless otherwise indicated. Statistical Endocrinol. Metab. 10,18–23 (1999). analyses were performed using GraphPad Prism 6 or 7 software. *p < 0.05, **p < 25. Ito, T., Hamazaki, Y., Takaori-Kondo, A. & Minato, N. Bone marrow 0.01, ***p < 0.001, ****p < 0.0001. endothelial cells induce immature and mature B cell egress in response to erythropoietin. Cell Struct. Funct. 42, 149–157 (2017). Data availability. RNA sequencing data have been deposited in the Gene 26. Heinrich, A. C., Pelanda, R. & Klingmuller, U. A mouse model for visualization Expression Omnibus (GSE104701). All data supporting the conclusions are and conditional mutations in the erythroid lineage. Blood 104, 659–666 (2004). available from the authors upon request. 27. Rajantie, I. et al. Bmx tyrosine kinase has a redundant function downstream of angiopoietin and vascular endothelial growth factor receptors in arterial endothelium. Mol. Cell. Biol. 21, 4647–4655 (2001). Received: 17 October 2017 Accepted: 12 May 2018 28. Ehling, M., Adams, S., Benedito, R. & Adams, R. H. Notch controls retinal blood vessel maturation and quiescence. Development 140, 3051–3061 (2013). 29. Kobayashi, H. et al. Angiocrine factors from Akt-activated endothelial cells balance self-renewal and differentiation of haematopoietic stem cells. Nat. Cell Biol. 12, 1046–1056 (2010). 30. Butler, J. M. et al. Endothelial cells are essential for the self-renewal and References repopulation of Notch-dependent hematopoietic stem cells. Cell Stem Cell 6, 1. Kfoury, Y. & Scadden, D. T. Mesenchymal cell contributions to the stem cell 251–264 (2010). niche. Cell Stem Cell 16, 239–253 (2015). 31. Poulos, M. G. et al. Endothelial Jagged-1 is necessary for homeostatic and 2. Gao, X., Xu, C., Asada, N. & Frenette, P.S. The hematopoietic stem cell niche: regenerative hematopoiesis. Cell Rep. 4, 1022–1034 (2013). from embryo to adult. Development 145, dev139691 (2018). 32. Itkin, T. et al. Distinct bone marrow blood vessels differentially regulate 3. Wei, Q. & Frenette, P. S. Niches for hematopoietic stem cells and their haematopoiesis. Nature 532, 323–328 (2016). progeny. Immunity 48, 632–648 (2018). 33. Bruns, I. et al. Megakaryocytes regulate hematopoietic stem cell quiescence 4. Kiel, M. J. et al. SLAM family receptors distinguish hematopoietic stem and through CXCL4 secretion. Nat. Med. 20, 1315–1320 (2014). progenitor cells and reveal endothelial niches for stem cells. Cell 121, 34. Nakamura-Ishizu, A., Takubo, K., Fujioka, M. & Suda, T. Megakaryocytes are 1109–1121 (2005). essential for HSC quiescence through the production of thrombopoietin. 5. Mendez-Ferrer, S. et al. Mesenchymal and haematopoietic stem cells form a Biochem. Biophys. Res. Commun. 454, 353–357 (2014). unique bone marrow niche. Nature 466, 829–834 (2010). 35. Zhao, M. et al. Megakaryocytes maintain homeostatic quiescence and promote 6. Omatsu, Y. et al. The essential functions of adipo-osteogenic progenitors as post-injury regeneration of hematopoietic stem cells. Nat. Med. 20, 1321–1326 the hematopoietic stem and progenitor cell niche. Immunity 33, 387–399 (2014). (2010). 36. Acar, M. et al. Deep imaging of bone marrow shows non-dividing stem cells 7. Ding, L., Saunders, T. L., Enikolopov, G. & Morrison, S. J. Endothelial and are mainly perisinusoidal. Nature 526, 126–130 (2015). perivascular cells maintain haematopoietic stem cells. Nature 481, 457–462 37. Pinho, S. et al. Lineage-biased hematopoietic stem cells are regulated by (2012). distinct niches. Dev. Cell 44, 634–641 (2018). 8. Ding, L. & Morrison, S. J. Haematopoietic stem cells and early lymphoid 38. Mignone, J. L., Kukekov, V., Chiang, A. S., Steindler, D. & Enikolopov, G. progenitors occupy distinct bone marrow niches. Nature 495, 231–235 (2013). Neural stem and progenitor cells in nestin-GFP transgenic mice. J. Comp. 9. Greenbaum, A. et al. CXCL12 in early mesenchymal progenitors is required Neurol. 469, 311–324 (2004). for haematopoietic stem-cell maintenance. Nature 495, 227–230 (2013). 39. Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 10. Asada, N. et al. Differential cytokine contributions of perivascular 15–21 (2013). haematopoietic stem cell niches. Nat. Cell Biol. 19, 214–223 (2017). 40. Anders, S., Pyl, P. T. & Huber, W. HTSeq—a Python framework to work with 11. Kunisaki, Y. et al. Arteriolar niches maintain haematopoietic stem cell high-throughput sequencing data. Bioinformatics 31, 166–169 (2015). quiescence. Nature 502, 637–643 (2013). 12. Sasine, J. P., Yeo, K. T. & Chute, J. P. Concise review: paracrine functions of vascular niche cells in regulating hematopoietic stem cell fate. Stem Cells Acknowledgements Transl. Med. 6, 482–489 (2017). We thank C. Prophete and P. Ciero for technical assistance and L. Tesfa, Y. Wang, and 13. Kusumbe, A. P., Ramasamy, S. K. & Adams, R. H. Coupling of angiogenesis D. Sun for help with cell sorting. We thank Drs. R. Adams, T. Nagasawa, and S. Morrison and osteogenesis by a specific vessel subtype in bone. Nature 507, 323–328 for providing genetically modified mice. This work was supported by R01 grants from the (2014). National Institutes of Health (NIH) (DK056638, HL069438, DK116312, DK112976 to P. 14. Ramasamy, S. K., Kusumbe, A. P., Wang, L. & Adams, R. H. Endothelial S.F.). We are also grateful to the New York State Department of Health (NYSTEM Notch activity promotes angiogenesis and osteogenesis in bone. Nature 507, Program) for shared facility (C029154) and research support (N13G-262) and the 376–380 (2014). Leukemia and Lymphoma Society’s Translational Research Program. 15. Kusumbe, A. P. et al. Age-dependent modulation of vascular niches for – haematopoietic stem cells. Nature 532, 380 384 (2016). Author contributions 16. Nombela-Arrieta, C. et al. Quantitative imaging of haematopoietic stem and C.X. designed and performed experiments, analysed and interpreted data, and wrote the progenitor cell localization and hypoxic status in the bone marrow manuscript. X.G., Q.W., and F.N. performed experiments. S.E.Z. and J.M. did the ana- microenvironment. Nat. Cell Biol. 15, 533–543 (2013). lysis of the RNAseq data. P.S.F. designed and supervised the study, interpreted data, and 17. Hooper, A. T. et al. Engraftment and reconstitution of hematopoiesis is wrote the manuscript. dependent on VEGFR2-mediated regeneration of sinusoidal endothelial cells. Cell Stem Cell 4, 263–274 (2009). 18. Schacht, V. et al. T1alpha/podoplanin deficiency disrupts normal lymphatic Additional information vasculature formation and causes lymphedema. EMBO J. 22, 3546–3556 Supplementary Information accompanies this paper at https://doi.org/10.1038/s41467- (2003). 018-04726-3. 19. Herzog, B. H. et al. Podoplanin maintains high endothelial venule integrity by interacting with platelet CLEC-2. Nature 502, 105–109 (2013). Competing interests: The authors declare no competing interests.

12 NATURE COMMUNICATIONS | (2018) 9:2449 | DOI: 10.1038/s41467-018-04726-3 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04726-3 ARTICLE

Reprints and permission information is available online at http://npg.nature.com/ appropriate credit to the original author(s) and the source, provide a link to the Creative reprintsandpermissions/ Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in indicated otherwise in a credit line to the material. If material is not included in the published maps and institutional afﬁliations. article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give © The Author(s) 2018

NATURE COMMUNICATIONS | (2018) 9:2449 | DOI: 10.1038/s41467-018-04726-3 | www.nature.com/naturecommunications 13